Chrome spinels occur in olivine rich cumulates in the Jinchuan intrusion, host to one of the largest known accumulations of magmatic Ni-Cu in the world. The chemistry and mode of occurrence of these spinels is compared with those in two other similar but unmineralized mafic-ultramafic intrusions in the same belt. The Jinchuan intrusion contains Mg-poor chromites with widely varying TiO2 contents, showing many features that are typical of chromites in normal tholeiitic layered intrusions. Wide compositional variance occurs on the scale of single thin sections, and is attributed to reaction between cumulus chromite and trapped inter-cumulus liquid. This process operated extensively, even on grains that were armored by cumulus olivine crystals. Different reaction paths correlate with the abundance of sulfides in the rock. The unmineralized intrusions show similar reaction trends, but are offset from the Jinchuan trends. Yejili spinels are generally higher in Al and Ga, while Zangbutai chromites are systematically higher in Fe3+. Jinchuan spinel-olivine pairs record equilibration temperatures well below the solidus, and are slightly but significantly lower than blocking temperatures recorded at Yejili. The Jinchuan ore zone contains low Cr, high Fe3+ chromites that are unusual and distinctive. A plausible model for these unusual compositions is that these grains originated as normal aluminous spinels crystallizing from the parent silicate melt, and subsequently became enlarged and modified during crystallization of the sulfide melt. Ti-enriched chromites enclosed within sulfides record a prehistory of extensive reaction with trapped liquid. Sulfide-associated chromites from Jinchuan are strongly nickel depleted relative to expected Ni values for their Fe3+ content when compared with the barren intrusions. This feature, combined with the distinctive chemistry of the sulfide-related spinels, may have applications in the use of heavy resistate minerals during exploration. Distinctively high Ti chromites are also a feature of intrusions associated with the Karoo flood basalts, suggesting that Jinchuan may have a similar affinity. However, TiO2 enrichment in chromite can be a product of interaction with differentiated trapped liquid, as indicated by similar trends in lava lakes, layered intrusions, and at Sudbury, and therefore cannot be taken to prove unusually Ti-rich parent magmas.